Veo 3
2025-05-20 · Publisher : Google · Domain : Artificial Intelligence
Google DeepMind unveils Veo 3, its video generation model that produces audio together with the image for the first time: ambient sounds, sound effects and synchronized dialogue, with better adherence to physics.
What changes
- Video generation with native audio: ambience, effects, dialogue and lip synchronization.
- Better quality than Veo 2, understanding of short scenarios written in the prompt, more realistic physics.
- Launched at I/O for Ultra subscribers in the United States (Gemini app and Flow tool) and for businesses on Vertex AI.
- On July 17, 2025, opened in paid preview in the Gemini API; Fast version on July 31.
- On September 8, 2025, moved to general availability with vertical 9:16 output, 1080p and a price reduction.
### Specifications - Identifiers: `veo-3.0-generate-preview` (July), then `veo-3.0-generate-001` and `veo-3.0-fast-generate-001` (stable). - Availability: Gemini app, Flow, Gemini API, Vertex AI. - Outputs marked with the SynthID watermark. ### Pricing | Version | USD per second (audio included) | |---|---| | Veo 3, API preview (July) | 0.75 | | Veo 3, general availability (September 8) | 0.40 | | Veo 3 Fast, preview (July, earlier pricing) | 0.40 | | Veo 3 Fast, general availability | 0.15 | ### Benchmarks No numerical score published in the announcements. ### Key takeaways Google's first consumer video model to include sound. Cutting the price by nearly half in September makes it practical for marketing and prototyping uses.
Links
Tags : Vidéo, Google, DeepMind, Veo, Audio